VERA: Variational Inference Framework for Jailbreaking Large Language Models

Published in Advances in Neural Information Processing Systems (NeurIPS 2025), 2025